Nm23-H2 Interacts with a G Protein-coupled Receptor to Regulate Its Endocytosis through an Rac1-dependent Mechanism*

G protein-coupled receptors (GPCRs) represent a vast family of transmembrane proteins involved in the regulation of several physiological responses. The thromboxane A2 receptor (present as two isoforms: TP (cid:1) and TP (cid:2) ) is a GPCR displaying diverse pharmacological effects. As seen for many other GPCRs, TP (cid:2) is regulated by agonist-induced internalization. In the present study, we report the identification by yeast two-hybrid screening of Nm23-H2, a nucleoside diphosphate kinase, as a new interacting molecular partner with the C-terminal tail of TP (cid:2) . This interaction was confirmed in a cellular context when Nm23-H2 was co-immunoprecipitated with TP (cid:2) in HEK293 cells, a process dependent on agonist stimulation of the receptor. We observed that agonist-induced internalization of TP (cid:2) was regulated by Nm23-H2 through modulation of Rac1 signaling. Immunofluorescence microscopy in HEK293 cells revealed that Nm23-H2

G protein-coupled receptors (GPCRs) 1 constitute the largest family of transmembrane proteins involved in signal transduction. They regulate a wide variety of physiological responses such as neurotransmission, inflammation, cell growth and differentiation, and smell and taste perception. For example, the thromboxane A2 receptor (TP) is implicated in the regulation of diverse pharmacological events, including platelet aggregation, constriction of vascular and bronchiolar smooth muscle cells, as well as mitogenesis and hypertrophy of vascular smooth mus-cle cells (1). Two TP receptor isoforms were identified that are generated by the alternative splicing of a single gene, TP␣ (343 amino acids) and TP␤ (407 amino acids), which share the first 328 amino acids (2,3). Previous experiments performed by Parent et al. (4) demonstrated that only TP␤, but not TP␣, undergoes agonist-induced and tonic (constitutive) internalization, which are dictated by distinct motifs in the C terminus of the TP␤ receptor. Our efforts in the laboratory have been focused on understanding the molecular mechanisms involved in the regulation of TP␤ (5,6). The majority of GPCRs undergo internalization following agonist stimulation. This internalization can participate in receptor desensitization, resensitization, degradation, or activation of signaling cascades such as mitogen-activated protein kinase pathways (7)(8)(9). It has long been questioned if GPCRs signaling and endocytosis constitute two separate events, and if they are regulated by distinct molecular mechanisms (9). In this regard, we have recently shown that G␣ q signaling is directly involved in agonist-induced internalization of the TP␤ receptor, suggesting that, in the case of some GPCRs, signaling and endocytosis are tightly connected (6). However, the molecular mechanism by which G␣ q protein regulates TP␤ agonist-induced internalization is still unclear and is one of the major interests of our laboratory. Growing evidence has shown that endocytosis of GPCRs is governed by signaling molecules (6,9). Indeed, some studies have demonstrated that members of the Rho family such as RhoA, Rac, and Cdc42 as well as a member of the ARF family, ARF6, play a crucial role in GPCR endocytosis (10 -12). Nevertheless, the regulation of these small G proteins and how they are involved in this process is not well understood. It has been suggested that the regulation of Rho/ARF-mediated actin cytoskeleton rearrangement is either directly or indirectly involved in the regulation of GPCR endocytosis (13)(14)(15)(16)(17)(18)(19)(20). In addition, several other signaling proteins participating in endocytosis have been identified. Recently, it has been shown that Nm23-H1 regulates dynamin-dependent endocytosis (12,21,22). Nm23 (also known as NDPK, NDK, and Awd) was first identified as a putative tumor metastasis suppressor and as an essential element of fly development (23). In human, Nm23 represents a family of nucleoside diphosphate kinases (NDPKs) encoded by eight different genes. According to their genomic architecture and phosphotransferase activity, human Nm23 genes were classified into two groups. Group I is represented by the Nm23-H1 to Nm23-H4 isoforms, whereas group II contains the Nm23-H5 to Nm23-H8 isoforms (23). Group I Nm23 proteins possess a highly conserved kinase active site. On the other hand there is no strict conservation of the kinase site motifs for group II, and only Nm23-H6 has been reported to display kinase activity (23). Nm23 proteins are multifunctional and were shown to be involved in a fascinating variety of cellular * This work was supported in part by a grant (to J.-L. P.) from the Canadian Institutes of Health Research (CIHR). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
§ Supported by a doctoral fellowship from the CIHR. ʈ Recipient of a New Investigator Award from the CIHR. To whom correspondence should be addressed: Service de Rhumatologie, Faculté de Médecine, Université de Sherbrooke, 3001, 12 e Ave. Nord, Fleurimont, Québec J1H 5N4, Canada. Tel.: 819-564-5264; Fax: 819-564-5265; E-mail: jean-luc.parent@USherbrooke.ca. 1 The abbreviations used are: GPCR, G protein-coupled receptor; TP, thromboxane A2 receptor; NDPK, nucleoside diphosphate kinase; GEF, guanine nucleotide exchange factor; TP␤CT, TP␤ C-terminal; ELISA, enzyme-linked immunosorbent assay; PBS, phosphate-buffered saline; TBS, Tris-buffered saline; BSA, bovine serum albumin; HA, hemagglutinin; GST, glutathione S-transferase; PBD, p21-binding domain; DP, prostaglandin in D 2 receptor; IP, prostacyclin receptor. activities, including proliferation, development, and differentiation (24). Several studies demonstrated that regulation of these processes by Nm23 proteins is the result of their ability to modulate diverse transmembrane signaling pathways, including transforming growth factor-␤1, nerve growth factor, platelet-derived growth factor, and insulin-like growth factor-1 as reviewed by Otero et al. (25). NDPK␤ (Nm23-H1) was recently shown to promote activation of the G␣ s subunit of heterotrimeric G proteins (25). Nm23 was also suggested to function as a nucleoside diphosphate kinase involved in the activation of dynamin by promoting its GTP loading (22). Inactivation of the Nm23-H1 (also known as awd) gene in Drosophila led to the inhibition of endocytosis occurring at nerve terminals (22). Moreover, Palacios et al. (12) reported that ARF6-GTP recruits Nm23-H1 to facilitate dynamin-mediated endocytosis of E-cadherin and transferrin receptors during disassembly of adherens junctions. They also proposed that Nm23-H1 promotes adherens junctions disassembly by inactivating the Rac1 signaling pathway. Furthermore, these studies demonstrated that Nm23-H1 inactivates Rac1 by binding Tiam1, a guanine nucleotide exchange factor (GEF) for Rac1 (12,27). In addition, the same authors showed that expression of Nm23-H1-⌬Kpn, a mutant that lacks the killer of prune homology domain involved in the oligomerization and the function of the wild-type Nm23, could reverse ARF6-mediated inactivation of Rac1 signaling, because it could not bind Tiam1 (12). To our knowledge, Nm23 proteins were never demonstrated to interact with a transmembrane receptor and more particularly, their role in GPCR endocytosis remains to be addressed.
In our attempt to identify putative factors involved in TP␤ agonist-mediated endocytosis, we performed yeast two-hybrid screening using a HeLa cell cDNA library. Interestingly, we identified Nm23-H2 as a protein capable of interacting with the C-terminal tail of TP␤. Here we report a novel interaction between Nm23-H2 and TP␤ that regulates TP␤ agonist-induced internalization through a Rac1 mechanism.
Two-hybrid Screening Assay-A yeast two-hybrid screening was performed following the two-hybrid system standard protocol (28). Briefly, the pAS2-1-TP␤CT plasmid was transformed into the yeast strain pJ69-4␣ according to the lithium yeast transformation protocol (28) generating a stably pAS2-1/CT-transformed yeast clone. This clone was then transformed with a Human HeLa MATCHMAKER cDNA Library or with the empty pGad424 plasmid (Clontech) (negative control) and grown on (Trp Ϫ , Leu Ϫ , and His Ϫ ) restricted media. Several clones showing positive interactions were then isolated, and these interactions were confirmed by growth on quadruple selective (Trp Ϫ , Leu Ϫ , His Ϫ , and Ade Ϫ ) media. pGAD-GH plasmids containing the library inserts from positive colonies were isolated and transformed into the DH10B bacterial strain. Plasmids were extracted from DH10B cells and transformed once more into yeast with either the bait (pAS2-1/TP␤CT) or the negative control (pAS2-1) and plated on quadruple selective medium (Trp Ϫ , Leu Ϫ , His Ϫ , and Ade Ϫ ) to confirm the interaction. The selected plasmids were then sequenced by dideoxy sequencing. The sequences obtained were searched by using the NCBI blast alignment tool (available at www.ncbi.nlm.nih.gov).
Cell Culture and Transfection-HEK293 cells were maintained in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal bovine serum (ICN) at 37°C in a humidified atmosphere containing 5% CO 2 . Transient transfections of HEK293 cells grown to 75-90% confluence were performed using FuGENE 6 TM (Roche Applied Science) according to the manufacturer's instructions. Empty pcDNA3 vector was added to keep the total DNA amount added per plate constant. Cells grown on 60-mm plates were transfected using 6 g of total DNA vectors.
Receptor Cell Surface Expression and Internalization Assays-Cell surface internalization of TP␤ and ␤ 2 AR were measured by ELISA using transiently transfected HEK293 cells as described previously (4). Briefly, 1.2 ϫ 10 6 cells were grown overnight in 60-mm plates. The cells were then transfected with pcDNA3-HA-TP␤ or other indicated receptor constructs combined with empty pcDNA3, pcDNA3-Nm23-H2 WT , pcDNA3-Nm23-H2 H118C , or pcDNA3-Nm23-H2-⌬Kpn. Transfected cells were maintained as described above for 24 h. Thereafter, 2 ϫ 10 5 cells were transferred to 24-well plates precoated with 0.1 mg/ml poly-Llysine (Sigma) and maintained for an additional 24 h. To assess the agonist-induced internalization, the transfected cells were washed once with phosphate-buffered saline (PBS) followed by stimulation with 100 nM U46619 at 37°C for the indicated times in Dulbecco's modified Eagle's medium. Thereafter, the cells were fixed with 3.7% formaldehyde plus TBS for 5 min at room temperature followed by three times wash with TBS. Nonspecific binding was blocked with TBS containing 1% BSA for 45 min at room temperature. The cells were then incubated with a HA-or FLAG-specific monoclonal antibody (BAbCO) at a dilution of 1:1000 in TBS/BSA for 1 h at room temperature. Three washes with TBS buffer followed, and cells were briefly reblocked for 15 min at room temperature. The cells were incubated with a 1:1000 in TBS/BSA of alkaline phosphatase-conjugated goat anti-mouse antibody (Bio-Rad) for 1 h at room temperature. The cells were washed three times with TBS, and a colorimetric alkaline phosphatase substrate (Bio-Rad) was added following the instructions from the manufacturer. The resulting colorimetric reactions were measured using a Titertek MultisKan MCC/ 340 spectrophotometer. Cells transfected with pcDNA3 were studied concurrently to determine background. Values shown represent the mean Ϯ S.E. of four to six independent experiments performed in triplicate.
Immunoprecipitations-We transfected 6-well plates of HEK293 cells with pcDNA3, pcDNA3-HA-TP␤, pcDNA3-Myc-Nm23-H2, pcDNA3-Myc-Nm23-H2 H118C , and pcDNA3-Myc-Nm23-H2-⌬Kpn in the different combinations indicated under "Results." Transfected cells were maintained as described above for 48 h and incubated for 0 -30 min at 37°C in the presence of 100 nM U46619 prior to harvesting. Thereafter, the cells were rinsed with ice-cold PBS and harvested in 800 l of lysis buffer (150 mM NaCl, 50 mM Tris, pH 8, 1% Nonidet P-40, 0.5% deoxycholate, 0.1% SDS, 10 mM Na 4 P 2 O 7 , 5 mM EDTA) supplemented with protease inhibitors (9 nM pepstatin, 9 nM antipain, 10 nM leupeptin, and 10 nM chymostatin) (Sigma). Following 60-min incubation of the cells in lysis buffer at 4°C with rotation, the lysates were clarified by centrifugation for 20 min at 14,000 rpm at 4°C. One to four micrograms of HA-specific monoclonal were added to the supernatant. After 60 min of incubation at 4°C, 50 l of 50% protein A-agarose pre-equilibrated in lysis buffer was added, followed by an overnight incubation at 4°C. Samples were then centrifuged for 1 min in a microcentrifuge and washed three times with 1 ml of lysis buffer. Immunoprecipitated proteins were eluted by addition of 50 l of SDS sample buffer followed by a 30-min incubation at room temperature. Initial lysates and immunoprecipitated proteins were analyzed by SDS-PAGE and immunoblotting performed by specific antibodies.
Recombinant Protein Production and Binding Assay-The cDNA encoding for TP␤CT (amino acid region 312-407 representing the whole C-terminal domain of TP␤) was subcloned in the pGEX-4-T1 vector (Amersham Biosciences), and the construct was used to produce a GST-tagged TP␤CT fusion protein in Escherichia coli BL21 by following the manufacturer's instruction. The recombinant TP␤CT was purified using glutathione-Sepharose TM 4B (Amersham Biosciences) as indicated by the manufacturer. The purified recombinant TP␤CT was analyzed by SDS-PAGE and was immunoblotted by using a GST-specific polyclonal antibody (Bethyl Laboratories). Glutathione-Sepharosebound GST-TP␤CT was incubated with Myc-NM23-H2 cellular extracts in binding buffer (10 mM Tris, pH 7.4, 150 mM NaCl, 1 mM EDTA, 10% glycerol, 0.5% Igepal) supplemented with proteases inhibitors (9 mM pepstatin, 9 mM antipain, 10 mM leupeptin, and 10 mM chymostatin) (Sigma). The binding reactions were then washed three times with binding buffer. SDS sample buffer was added to the binding reactions, and the tubes were boiled for 5 min. The binding reactions were analyzed by SDS-PAGE, and immunoblotting was performed with the indicated specific antibodies.
Inositol Phosphates Measurement in Cells-Inositol phosphates measurements were performed as described previously (4). 2 ϫ 10 5 HEK293 cells were grown overnight in 12-well plates. The cells were then co-transfected as described above with the indicated constructs. The cells were labeled the following day for 18 -24 h with 4 Ci/ml myo-[ 3 H]inositol (Amersham Biosciences) in Dulbecco's modified Eagle's medium (high glucose, without inositol). The cells were washed once in phosphate-buffered saline and incubated in pre-warmed Dulbecco's modified Eagle's medium (high glucose, without inositol) supplemented with 0.5% bovine serum albumin, 20 mM Hepes, pH 7.5, and 20 mM LiCl for 10 min. After a 30-min stimulation with 500 nM U46619, the medium was removed, and the reactions were terminated by addition of 0.8 ml of 0.4 M chilled perchloric acid. Cells were then collected in microcentrifuge tubes, and a 0.5 volume of a 0.72 N KOH 0.6, M KOHCO 3 solution was added. Tubes were mixed and centrifuged for 5 min at 14,000 rpm in a microcentrifuge. Inositol phosphates were separated on Bio-Rad AG 1-X8 columns and eluted with 1.5 M ammonium formate. Total labeled inositol phosphates were then measured by liquid scintillation counting.
Immunofluorescence Microscopy-TP␤ and Nm23-H2 intracellular localization was assessed by immunofluorescence microscopy. 1.2 ϫ 10 6 HEK293 cells were grown overnight in 60-mm plates as described above. The cells were then transfected with an empty pcDNA3 vector or co-transfected with pcDNA3-HA-TP␤ and pcDNA3-Myc-Nm23-H2, and maintained overnight as described above. Then 2 ϫ 10 5 cells were transferred on coverslips and further grown overnight. The cells were then stimulated at 37°C with 100 nM U46619 before being fixed with 3% paraformaldehyde plus PBS for 30 min at room temperature. Thereafter, the cells were washed with PBS and permeabilized with 0.1% Triton X-100 plus PBS for 10 min at room temperature. Nonspecific binding was blocked with 0.1% Triton X-100 plus PBS containing 5% nonfat dry milk for 30 min at room temperature. The cells were then incubated with an HA-specific monoclonal (1:500 dilution) and Mycspecific polyclonal (1:200 dilution) antibodies (1:500 dilution) for 1 h at room temperature in PBS supplemented with 5% nonfat dry milk. Then, cells were washed three times with 0.1% Triton X-100 plus PBS, followed by incubation with a goat anti-mouse fluorescein isothiocyanate-conjugated and Texas Red goat anti-rabbit secondary antibodies (Molecular Probes) at a dilution of 1:200 for 1 h at room temperature. The cells were washed six times with permeabilization buffer. Finally, coverslips were mounted using Vectashield mounting medium (Vector Laboratories) and examined by immunofluorescence microscopy on a Nikon Eclipse TE2000-U microscope using a 60ϫ objective. Images were collected using SimplePCI Camera software and processed with Adobe Photoshop software.
Rac1 Activation Assay-Rac1 activation was assessed by the PBD assay. This method relies on the capacity of the active form of Rac1 to bind the N-terminal regulatory region of (p21)-activated kinase PAK1 referred to as the p21-binding domain (PBD) (29,32). Briefly, 1.8 ϫ 10 6 HEK293 cells were grown overnight in 100-mm plates. The cells were then co-transfected as described above with the indicated constructs. Transfected cells were maintained as described above for 48 h and incubated for 0 -120 min at 37°C in the presence of 100 nM U46619 prior to harvesting. Thereafter, the cells were rinsed with ice-cold PBS and harvested in 800 l of lysis buffer (200 mM NaCl, 50 mM Tris-HCl, pH 7.5, 10 mM MgCl 2 , 0.5% Triton X-100, 5% glycerol, 1 mM phenylmethylsulfonyl fluoride) supplemented with protease inhibitors (9 nM pepstatin, 9 nM antipain, 10 nM leupeptin, and 10 nM chymostatin) (Sigma). Cell extracts (10 g) were incubated with glutathione-Sepharose-bound GST-PBD (a kind gift of Gary Bokoch, The Scripps Research Institute, La Jolla, CA) for 1 h at 4°C. The binding reactions were then centrifuged for 2 min at 2000 rpm, and the bead pellets were washed three times with washing buffer A (25 mM Tris-HCl, pH 7.6, 1 mM dithiothreitol, 30 mM MgCl 2 , 40 mM NaCl, 0.5% Triton, plus antiproteases) followed by two washes in buffer B (25 mM Tris-HCl, pH 7.6, 1 mM dithiothreitol, 30 mM MgCl 2 , 40 mM NaCl, plus antiproteases). 20 l of Laemmli sample buffer was then added to the bead pellet. The binding reactions were then analyzed by immunoblotting performed by the use of Rac1-speific monoclonal antibody (BD Biosciences).

RESULTS
Nm23-H2 Interacts with TP␤-We have recently shown that the C terminus of TP␤ confers specific regulatory mechanisms to this isoform of the thromboxane A 2 receptor (4,26). To identify putative proteins, that could interact with the C terminus of TP␤, we performed yeast two-hybrid screening experiments using the yeast strain pJ69-4␣ transformed with pAS2.1-TP␤CT and a human HeLa cell cDNA library. A total of 1.5 ϫ 10 6 independent clones were screened yielding over 300 positives. One hundred clones demonstrating a strong growth on selective yeast medium (Trp Ϫ , Leu Ϫ , His Ϫ , and Ade Ϫ ) were then isolated and characterized by dideoxy sequencing and then aligned using the NCBI blast alignment search tool. Three clones contained a cDNA fragment coding for the Nm23-H2 protein. Other members of the Nm23 protein family were not isolated in the clones that were sequenced. In this study we report the characterization of Nm23-H2 as a putative protein, which strongly interacts with the C-terminal tail of TP␤. As shown in Fig. 1A, strong growth on Trp Ϫ , Leu Ϫ , His Ϫ , and Ade Ϫ medium was present only in yeast transformed with pAS2.1-TP␤CT and pGADGH-Nm23-H2, indicating that Nm23-H2 is interacting with the C terminus of TP␤.
Co-immunoprecipitations and in Vitro Binding Assay-To investigate the interaction of Nm23-H2 with TP␤ in a cellular context, we performed immunoprecipitation experiments in HEK293 cells transfected with pcDNA3-Myc-Nm23-H2 and pcDNA3-HA-TP␤ in the presence or absence of U46619 stimulation. Cell lysates were incubated with a HA-specific monoclonal antibody and protein A-agarose, and immunoprecipitation reactions were analyzed by immunoblotting with a Myc-specific polyclonal antibody. Our results demonstrate that Myc-Nm23-H2 was co-immunoprecipitated with TP␤ following agonist treatment (Fig. 1B). This suggests that the interaction of Nm23-H2 with TP␤ is modulated by agonist stimulation. Moreover, to further confirm the interaction of Nm23-H2 with the C-terminal tail of TP␤, we performed an in vitro binding assay using the purified recombinant C-terminal tail of TP␤ fused to GST (GST-TP␤CT) along with HEK293 cell extracts expressing Myc-Nm23-H2. The results obtained showed that Nm23-H2 could bind specifically to glutathione-Sepharosebound GST-TP␤CT and not to glutathione-Sepharose-bound GST (Fig. 1C). Taken together, our data demonstrate that Nm23-H2 interacts directly with the C-terminal tail of TP␤ in an agonist-dependent manner. Interestingly, this constitutes the first study whereby a direct interaction of an Nm23 protein with a G protein-coupled receptor, and to our knowledge to any transmembrane receptor, is shown.
Nm23-H2 Inhibits TP␤ Agonist-induced Internalization-We were then interested in determining the role of Nm23-H2 in TP␤ regulation. Nm23-H1 and Nm23-H2 are 88% identical. Two dominant-negative mutants of Nm23-H1 have been reported. Nm23-H1 H118C was characterized as a kinasedefective mutant of Nm23-H1, whereas the Nm23-H1-⌬Kpn mutant was shown to be unable to inactivate Rac1 signaling, because it could not bind and sequester Tiam1 (an Rac1 GEF) (12). The same mutations were thus introduced into Nm23-H2. We first assessed how Nm23-H2 and its two mutants would regulate agonist-induced internalization of TP␤. Fig. 2A shows the results of a TP␤ internalization assay following a 2-h incubation with U46619 performed by ELISA (4) in HEK293 cells transfected with HA-TP␤ and either Myc-Nm23-H2 WT , Myc-Nm23-H2 H118C , Myc-Nm23-H2-⌬Kpn, or pcDNA3. Intriguingly, we observed that only Nm23-H2-⌬Kpn significantly inhibited the agonist-induced internalization of TP␤, whereas expression of Nm23-H2 WT or Nm23-H2 H118C had no apparent effect on the latter. A time course of receptor internalization Nm23-H2 Regulates TP␤ Agonist-induced Internalization was performed to further characterize the effect of Nm23-H2-⌬Kpn. Data found in Fig. 2B confirmed that only Nm23-H2-⌬Kpn significantly inhibited agonist-induced internalization of TP␤ with even more drastic effects at earlier time points. These data suggest that the kinase activity of Nm23-H2 is not involved in the process of agonist-induced internalization of TP␤. Failure of Nm23-H2 H118C to inhibit TP␤ internalization signi-fies that dynamin is not likely involved in modulation of this process by Nm23-H2. On the other hand, the killer of prune homology domain of Nm23-H2, deleted in Nm23-H2-⌬Kpn, seems to be important for the progress of the agonist-induced internalization of TP␤, indicating that Nm23-H2 is positively involved in this process. Furthermore, using similar experimental approaches, we determined that Nm23-H2-⌬Kpn did FIG. 1. The TP␤ C-terminal tail interacts with Nm23-H2. A, yeast two-hybrid screening was performed using TP␤CT and the Human HeLa MATCHMAKER cDNA Library. Nm23-H2 was characterized as a novel molecular partner that interacts with TP␤CT. This interaction was confirmed by the use of the selective yeast media (Trp Ϫ , Leu Ϫ , His Ϫ , and Ade Ϫ ). B, co-immunoprecipitation experiments were performed in HEK293 cells transfected with pcDNA3-Myc-Nm23-H2 and pcDNA3-HA-TP␤. The result illustrated here shows that Nm23-H2 co-immunoprecipitated with TP␤ in U46619-stimulated HEK293 cells. C, the binding assay was performed using purified recombinant GST-TP␤CT protein and Myc-Nm23-H2-expressing HEK293 cell extracts. Nm23-H2 binds to glutathione Sepharose-bound GST-TP␤CT but not to glutathione-Sepharosebound GST. Data presented are representative of three different experiments. not significantly inhibit agonist-induced internalization of the G␣ s -coupled DP and IP receptors, as well as of the predominantly G␣ i -coupled CRTH2 receptor (Fig. 3). Internalization of the ␤2-adrenergic receptor was increased by the Nm23-H2-⌬Kpn mutant. In contrast, Nm23-H2-⌬Kpn expression resulted in a ϳ50% inhibition of agonist-induced endocytosis of two other G␣ q -coupled receptors, the angiotensin II AT1 and the platelet-activating factor receptors (Fig. 3). This suggests that the role of Nm23-H2 in endocytosis is specific to some GPCRs. These results are in accordance with growing evidence from us and others showing that endocytosis of different GPCRs is regulated by distinct molecular events.
Nm23-H2 Does Not Regulate TP␤-mediated G␣ q Signaling-Nm23-H1 was shown to form a complex with the G␤␥ dimers and promote G protein activation by increasing the high energetic phosphate transfer to GDP (30). Kikkawa et al. (31) published that Nm23-H1 also enhances G␣ s signaling. It is the histidine kinase activity of Nm23-H1 that is involved in this phosphate transfer (31). Thus we were next interested in verifying if Nm23-H2 and its two mutants would alter TP␤ signaling. Our data indicated that none of the Nm23-H2 WT , Nm23-H2-⌬Kpn, or Nm23-H2 H118C constructs affected TP␤mediated G␣ q or G␣ s signaling as determined by measuring total inositol phosphates and cAMP generation (data not shown). Similar results were obtained with constitutive forms of G␣ q and G␣ s (data not shown). We previously demonstrated that G␣ q signaling was necessary for TP␤ internalization. The fact that Nm23-H2 H118C is not affecting G␣ q signaling concurs with its lack of effect on TP␤ agonist-induced internalization. Similarly, this also suggests that Nm23-H2-⌬Kpn inhibition of TP␤ agonist-induced internalization is accomplished by a mechanism that is not related the G␣ protein signaling. Rac1 Signaling Regulates TP␤ Endocytosis-Interestingly, Nm23-H1-⌬Kpn, lacking the killer of prune homology domain involved in the binding of Tiam1 (a Rac1 GEF), acts as a dominant-negative mutant toward the endogenous Nm23-H1 and as such interferes with the ARF6-mediated decrease of Rac1-GTP (12). Thus, we hypothesized that Nm23-H2 could modulate the agonist-induced internalization of TP␤ by mediating the inactivation of Rac1 signaling. However, to investigate this hypothesis, we first had to verify, on one hand, the effect of Rac1 signaling on the agonist-induced internalization of TP␤ and, on the other hand, the modulation of Rac-1 activation following TP␤ agonist stimulation. To this end, we studied the effect of expression of Rac1(V12), a constitutively active form of Rac1, on the agonist-induced internalization of TP␤. As seen in Fig. 4, Rac1(V12) significantly inhibited the agonistinduced internalization of TP␤. Thus, it appears that continuous activation of the Rac1 signaling pathway interfered with agonist-induced internalization of TP␤. Rac1(N17), a dominant-negative mutant of Rac1, showed no effect at 120 min of

Nm23-H2 Regulates TP␤ Agonist-induced Internalization
TP␤ agonist-induced internalization. This will be further addressed below. The activation state of Rac1 was determined by PBD pull-down assays on cell lysates of TP␤-expressing cells following a time-course stimulation of the receptor. The Nterminal regulatory region of (p21)-activated kinase PAK1 referred to as the p21-binding domain (PBD) binds specifically to the Rac1-GTP-activated form. Samples from the GST-PBD pull-down assays (32) were then analyzed by immunoblotting with a Rac1-specific monoclonal antibody. Fig. 5A shows that TP␤ stimulation produced a rapid activation of Rac1 at 15 min, which peaked at 30 -45 min, followed by a decrease in activated Rac1 at 60 min and complete Rac1 inactivation at 120 min. Noteworthy, the progressive inactivation of Rac1 after 60 min of stimulation coincides with the time period where the maximal agonist-induced internalization of TP␤ is observed (Fig. 2B and Ref. 4). Interestingly, expression of Nm23-H2-⌬Kpn resulted in a sustained activation of Rac1 over the time interval that we studied (Fig. 5B). This result suggests that Nm23-H2-⌬Kpn interferes with the inactivation of Rac1 mediated by an endogenous Nm23 protein. We then speculated that the lack of effect of the dominant-negative Rac1(N17) mutant in Fig. 4 was due to the time point at which the experiment was performed. Because there is no activated Rac1 at 120 min (Fig. 5A), then Rac1(N17) could not have any effect. A time course of TP␤ internalization in presence of Rac1(N17) over periods of time where we observed Rac1 activation was then carried out. As can be seen in Fig. 6, Rac1(N17) increased TP␤ internalization over the time frame of Rac1 activation by TP␤, indicating that blocking Rac1 signaling promotes internalization of the receptor.
Taken together, our findings show that activation of Rac1 signaling interferes with TP␤ agonist-induced internalization. Maximal receptor internalization occurs when Rac1 is inactivated. Our data also indicate that Nm23-H2-⌬Kpn inhibited TP␤ agonist-induced internalization through maintenance of activated Rac1 protein.
Rac1 Signaling Antagonizes the G␣ q -induced Internalization of TP␤ Receptor-We recently reported that G␣ q signaling is necessary in the agonist-induced internalization of TP␤. In this previous study, we have demonstrated that G␣ q -R183C (a constitutive form of G␣ q ) was sufficient to induce the internalization of TP␤ in absence of agonist stimulation (6). Here, we were thus interested in investigating the effect of the activated Rac1(V12) mutant on G␣ q -R183C-induced internalization of TP␤. In this experiment, G␣ q -R183C-induced internalization is assessed by measuring cell surface receptor expression by ELISA as we described previously (6). The results obtained show that G␣ q -R183C induced a drastic endocytosis of TP␤ receptor as its cell surface expression decreased significantly, as expected (6). However, Rac1(V12) expression interfered with G␣ q -R183C-induced internalization of TP␤ (Fig. 7). This suggests that Rac1 signaling antagonizes the function of G␣ q signaling in the process of TP␤ endocytosis.
TP␤-mediated Recruitment of Nm23-H2 to the Plasma Membrane-To assess the intracellular localization of Nm23-H2 and its regulation by TP␤ stimulation, we performed immunofluorescence microscopy using HEK293 cells transfected with pcDNA3-HA-TP␤ and pcDNA3-Myc-Nm23-H2. Using Myc-specific polyclonal and HA-specific monoclonal antibodies, double labeling of transfected cells was then performed. In the absence of TP␤ stimulation, Nm23-H2 has a predominantly cytoplasmic and nuclear distribution while TP␤ has both a cytoplasmic and membrane repartition (Fig. 8, upper panel). However, following U46619 stimulation, Nm23-H2 undergoes translocation to the plasma membrane resulting in its significant co-localization with TP␤ (Fig. 8, bottom panel). Co-localization of both proteins is also detected in intracellular compartments.

DISCUSSION
In the past few years, the relation between GPCRs signaling and their endocytosis has been the subject of debate. We have recently shown that the first step of regulation of TP␤ agonistmediated endocytosis involves the activation of G␣ q signaling (6). However, the exact molecular mechanisms and partners involved in this regulation are still unknown. To identify such molecular partners we performed yeast two-hybrid screening using the C-terminal tail of TP␤ and a human HeLa cell cDNA library. Here we report that Nm23-H2 interacts with the Cterminal tail of TP␤. Nm23, also known as NDP kinase (NDPK), is a 17-kDa histidine kinase involved in the regulation of a wide variety of cellular functions. Nm23 was shown to regulate cell growth, differentiation, tumor metastasis, kinase signal transduction, GTPase activation, and very recently endocytosis events (12,22,24). Nm23-H2 is one of the eight members of the Nm23 family. Co-immunoprecipitation results demonstrated that Nm23-H2 interacts with TP␤ in a cellular context. Interestingly, this interaction seemed to be triggered by the activation of TP␤, suggesting that Nm23-H2 could be involved in the regulation of the molecular events subsequent to TP␤ agonist stimulation. Thus we first investigated the effect of Nm23-H2 on TP␤ agonist-induced internalization. Coexpression of TP␤ and Nm23-H2 WT in HEK293 cells did not change the agonist-induced internalization of TP␤. Two dominant-negative mutants of Nm23-H1 are known in the literature (12). Nm23-H1 H118C is a kinase-deficient mutant shown to interfere with dynamin-mediated endocytosis (12). On the other hand, Nm23-H1-⌬Kpn, which lacks the killer of prune homology domain, is unable to bind Tiam1 (an Rac1 GEF) (12). The corresponding mutations were introduced into Nm23-H2, because Nm23-H1 and Nm23-H2 are highly identical (100% identity in the regions of the mutations). The ability of the two mutants to modulate TP␤ agonist-induced internalization was then evaluated. Nm23-H2 H118C had no effect on the agonistpromoted internalization of TP␤ suggesting that the kinase activity of Nm23-H2 is not involved in this process. We previously reported that TP␤ internalization is dynamin-dependent (4). In this case however, the Nm23-H2 H118C data indicate that Nm23-H2 would regulate TP␤ internalization through a mechanism not involving dynamin. Surprisingly, Nm23-H2-⌬Kpn expression significantly inhibited the agonist-mediated inter- nalization of TP␤ identifying the killer of prune homology domain of Nm23-H2 as a regulator of this process. There is specificity of Nm23-H2-⌬Kpn toward TP␤, because this mutant failed to prevent internalization of the G␣ s -coupled ␤ 2 -adrenergic, DP, and IP receptors. Internalization of the ␤ 2 -adrenergic receptor was increased by Nm23-H2-⌬Kpn expression, which could suggest that the presence of activated Rac1 promotes internalization of this receptor. Nm23-H2-⌬Kpn inhibited by ϳ50% the endocytosis of the G␣ q -coupled angiotensin II AT 1 and platelet-activating factor receptors. On the other hand, internalization of the CRTH2 receptor was not significantly affected by Nm23-H2-⌬Kpn expression. The CRTH2 receptor is predominantly a G␣ i -coupled receptor but was also reported to be coupled to G␣ q (33). Thus our data suggest that regulation of GPCR endocytosis by Nm23-H2 is a pathway specific to at least some, but not all, G␣ q -coupled receptors. Differences in the complement of proteins expressed or localized at particular subcellular sites (compartmentalization) between cell types could be a determinant of this pathway for a given receptor. More studies will be necessary to fully understand the speci-ficity of Nm23-H2 toward various classes of GPCRs. Our preliminary data indicate that agonist-induced endocytosis of TP␤ is not modulated by expression of Nm23-H1, Nm23-H1 H118C , or Nm23-H1-⌬Kpn constructs in our system (data not shown), further indicating specificity in the TP␤/Nm23-H2 mechanism. G␣ q signaling by TP␤ is necessary for the activation of the machinery responsible for internalization of this receptor, and interfering with this signaling pathway results in a drastic decrease in receptor internalization (6). We observed that Nm23-H2-⌬Kpn did not affect G␣ q signaling, as determined by inositol phosphate measurement. This indicated that Nm-23-H2-⌬Kpn was not exerting its effect at the level of G␣ q activity. Interestingly, Palacios et al. (12) recently published that ARF6, which is involved in membrane trafficking, facilitated the process of membrane endocytosis by recruiting Nm23-H1. The same study showed that the killer of prune homology domain of Nm23-H1 mediated the inactivation of Rac1 signaling by binding Tiam1 and interfering with its GEF activity on Rac1. Because our results demonstrated that Nm23-H2-⌬Kpn inhibited TP␤ agonist-induced internalization, we hypothesized that Nm23-H2 could play a positive role in TP␤ internalization by promoting the inactivation of Rac1 signaling. This hypothesis implied first that Rac1 signaling could interfere with TP␤ agonist-induced internalization and second that Rac1 signaling has to be inactivated for the internalization process to occur. Thus, on the first hand, we investigated the effect of Rac1 signaling on TP␤ internalization with the use of Rac1(V12), a constitutively active mutant of Rac1. This revealed that Rac1(V12) expression led to a significant inhibition of TP␤ agonist-induced internalization. This result suggested that the Rac1 signaling pathway negatively affected the agonist-induced internalization of TP␤. Rac1 activation was then assessed using a well characterized Rac1-GTP p21-binding domain assay (PBD assay) (32) to monitor the Rac1 activity during the process of agonist-induced internalization of TP␤. We have seen that TP␤ agonist stimulation promoted a tran-sient increase in the cellular amount of active Rac1-GTP, which peaked at 30 -45 min and steadily declined afterward. Interestingly, we demonstrated that, in the presence of Nm23-H2-⌬Kpn, a sustained activation of Rac1 was observed during TP␤ stimulation. This suggests that Nm23-H2 is involved in TP␤induced modulation of Rac1 signaling. These findings are in accordance with the hypothesis postulated above. It is interesting to note that constitutively active Rac1(V12) blocked TP␤ internalization and that maximal TP␤ internalization coincided with the moment when Rac1 was inactivated. This indicates that Rac1 activation by TP␤ would slow down its internalization, which would pick up once Rac1 gets inactivated. This statement is supported by the fact that Rac1(N17), a dominant-negative mutant of Rac1, promoted a significant increase in the rate of agonist-induced internalization of TP␤ at 30, 45, and 60 min, which coincides with the activation of Rac1 signaling. Rac1(N17) had no effect on TP␤ agonist-induced internalization at 120-min U46619 stimulation, a result that can be explained by the absence of Rac1-GTP at this time point. Furthermore, sustained activation of Rac1 signaling mediated by either Nm23-H2-⌬Kpn or Rac1(V12) expression led to a significant decrease in the agonist-induced internalization of TP␤. Its has been proposed that Rac1 could regulate receptor endocytosis by modulating phosphatidylinositol 4,5-bisphosphate turnover at the plasma membrane (10). In addition, Lamaze et al. (34) demonstrated that the active mutant of Rac1 blocked the internalization of the transferrin receptor. However, the molecular mechanism involved in this regulation is still unclear. Moreover, an increase in the localization of Nm23-H2 to the plasma membrane was shown following TP␤ receptor agonist stimulation by immunofluorescence microscopy suggesting that TP␤ recruited Nm23-H2 to the plasma membrane. This recruitment of Nm23-H2 is probably necessary for Tiam1 binding and inactivation of Rac1 signaling, both latter proteins being present at the plasma membrane (12,23,24).
Growing evidence demonstrates the crucial role played by GPCRs signaling in their endocytosis process (6,9). We recently reported that G␣ q signaling is strongly involved in the regulation of agonist-induced internalization of G␣ q -coupled GPCRs (6). Here we have shown that, contrary to G␣ q signaling, Rac1 activation interferes with the agonist internalization FIG. 7. Rac1 signaling interferes with G␣ q -induced internalization of TP␤. G␣ q -induced internalization of TP␤ was assessed by measuring the cell surface expression of the receptor by ELISA. Cell surface expression of TP␤ was compared between HA-TP␤-expressing HEK293 cells transfected with either pcDNA3, pcDNA3, and pcDNA3-G␣ q -R183C or pcDNA3-G␣ q -R183C and pHOOK-Rac1(V12).
FIG. 8. Immunofluorescence microscopy of TP␤ co-localization with Nm23-H2 in HEK293 cells. Immunofluorescence experiments were performed to determine the intracellular localization of Nm23-H2 and TP␤ using HEK293 cells transiently transected with pcDNA3-HA-TP␤ and pcDNA3-Myc-Nm23-H2 constructs. The transfected cells were incubated with 500 nM U46619 or not for 90 min. HA-specific monoclonal as well as Myc-specific polyclonal antibodies were used to label HA-TP␤ and Myc-Nm23-H2, respectively. U46619 stimulation promotes extensive colocalization of TP␤ with Nm23-H2 at both the cell membrane and in intracellular compartments. of TP␤. G␣ q and Rac1 signaling appears to play an antagonist role in the internalization process of TP␤. Indeed, Rac1(V12) inhibited G␣ q -R183C-induced internalization of the receptor. Rac1 signaling has been shown to regulate the actin cytoskeleton and membrane organization. It is therefore possible that Rac1 signaling activation leads to an actin cytoskeleton organization or a plasma membrane architecture not favorable for TP␤ endocytosis. In light of this study, one can imagine that agonist-induced internalization of TP␤ and perhaps other GPCRs involves a tight coordination of different signaling pathways.
Briefly, we reported here the first demonstration of an interaction of a member of the Nm23 proteins with a membrane receptor, in this particular case, a GPCR. This interaction regulates internalization of the receptor through a Rac1-dependent mechanism. Specificity was seen in the GPCRs affected, indicating that our findings constitute a novel, distinct molecular regulatory mechanism of GPCR internalization.